Natural language processing method, electronic device and computer program product

Abstract

The invention provides a natural language processing method, electronic equipment and a computer program product. The natural language processing method comprises the following steps: in response to a received service query in a natural language form, analyzing the service query through a first neural network model to obtain service parameter information; based on the first mapping relationship, determining a parameter conversion rule, a request template and an interface corresponding to the data source identifier in the service parameter information; based on a parameter conversion rule, converting the service parameter information into request parameter information conforming to an interface technical specification; injecting the request parameter information into a request template to generate a request; sending a request to the interface, wherein the request is used for triggering the interface to return response data; and generating reply content of the business query based on response data returned by the interface.

Description

Technical Field

[0001] This disclosure relates to a natural language processing method, an electronic device, a readable storage medium, and a computer program product. Background Technology

[0002] During the intelligent transformation of the real estate brokerage industry, staff need to regularly monitor and make decisions regarding the operational status of stores and market dynamics and risks within their jurisdiction.

[0003] In related technologies, enterprises typically set up multiple data sources internally to support operational tasks such as store alerts, regional evaluation report generation, and business performance monitoring. However, because each data source has its own proprietary interface technical specifications, involving different parameter structures and calling logic, staff need to manually write requests to access different interfaces, leading to low operational efficiency. Summary of the Invention

[0004] This disclosure provides a natural language processing method, an electronic device, a readable storage medium, and a computer program product.

[0005] According to one aspect of this disclosure, a natural language processing method is provided, comprising: in response to receiving a business query in natural language form, parsing the business query through a first neural network model to obtain business parameter information; determining, based on a first mapping relationship, a parameter conversion rule, a request template, and an interface corresponding to a data source identifier in the business parameter information; converting the business parameter information into request parameter information conforming to the interface technical specifications based on the parameter conversion rule; injecting the request parameter information into the request template to generate a request; sending the request to the interface, the request being used to trigger the interface to return response data; and generating a response content for the business query based on the response data returned by the interface.

[0006] Based on one technical solution, the differences in underlying data source interfaces are effectively shielded, achieving a unified entry point and automated process for accessing different data sources. This allows users to quickly perform various operational operations by inputting business queries in a unified manner using natural language, significantly lowering the user threshold and improving operational efficiency. Furthermore, because interface access requests are automatically generated through parameter conversion rules and request templates, even when adding new data sources or adjusting existing technical specifications, users do not need to memorize new interface technical specifications. Adding new data sources or adjusting technical specifications can be quickly achieved by supplementing or adjusting parameter conversion rules and request templates, resulting in greater scalability.

[0007] According to at least one embodiment of the natural language processing method of this disclosure, the business query is parsed using a first neural network model to obtain business parameter information, including: acquiring knowledge data of the domain to which the business query belongs; and parsing the business query using the first neural network model based on the knowledge data to obtain the business parameter information.

[0008] According to the technical solution of this embodiment, domain-related knowledge data is introduced during the parsing process, which enables the first neural network model to have a stronger semantic understanding ability. Especially when facing professional terms, polysemous words or ambiguous expressions, it can combine domain knowledge data to perform accurate disambiguation and parameter recognition, thereby improving the accuracy and coverage of business parameter information extraction.

[0009] According to at least one embodiment of the natural language processing method of this disclosure, the parameter conversion rule includes one or more of the following: parameter conversion based on a second mapping relationship, parameter conversion by calling a target interface, and conversion of relative time to absolute time.

[0010] According to the technical solution of this embodiment, different types of parameter conversion rules are defined to meet diverse parameter conversion needs. This modular and combinable rule design improves the flexibility and versatility of parameter conversion, enabling it to efficiently support parameter conversion in various business scenarios and reduce the workload of customized development.

[0011] According to a natural language processing method of at least one embodiment of this disclosure, the second mapping relationship includes at least a mapping relationship between business parameter names and request parameter names; based on the parameter conversion rule, converting the business parameter information into request parameter information conforming to the interface technical specification includes: based on the second mapping relationship, determining the target request parameter name corresponding to the target business parameter name in the business parameter information; using the target business parameter value corresponding to the target business parameter name as the target request parameter value corresponding to the target request parameter name; and using the target request parameter name and the target request parameter value corresponding to the target request parameter name as the request parameter in the request parameter information.

[0012] According to the technical solution of this embodiment, parameter conversion is performed based on the second mapping relationship, realizing automatic name mapping between business parameters and request parameters, and solving the problem of inconsistent field naming between the front-end natural language expression and the back-end system. Furthermore, by establishing a standardized second mapping relationship, new interfaces can be quickly adapted without modifying the core logic, thereby improving development efficiency and making it suitable for complex environments with multiple data sources.

[0013] According to at least one embodiment of the natural language processing method of this disclosure, based on the parameter conversion rule, the business parameter information is converted into request parameter information conforming to the interface technical specification, including: in response to the business parameter information including geographic region information and object type information, obtaining an object identifier conforming to the geographic region information and the object type information through a target interface call; and using the object identifier and the object type information as object parameters in the request parameter information.

[0014] According to the technical solution of this embodiment, by calling the target interface to perform parameter conversion, the corresponding object identifier can be automatically obtained when the business parameter information includes geographical area information and object type information. This enriches the object parameters in the request parameter information, avoids query failure due to the lack of accurate object identifiers, enhances the fault tolerance capability for fuzzy input and the completeness of data acquisition, and improves the query success rate.

[0015] According to at least one embodiment of the natural language processing method of this disclosure, based on the parameter conversion rule, the business parameter information is converted into request parameter information conforming to the interface technical specification, and the method further includes: in response to the business parameter information including a relative time range, determining the start and end absolute times corresponding to the relative time range based on the current absolute time, and determining the time granularity; and using the start and end absolute times and the time granularity as time parameters in the request parameter information.

[0016] According to the technical solution of this embodiment, the relative time range expressed in natural language is automatically converted into absolute time that conforms to the interface technical specifications, and the time granularity is supplemented, so that the time parameters in the request parameter information conform to the interface technical specifications, improving the accuracy and timeliness of time-related queries, while reducing the user's input burden.

[0017] According to at least one embodiment of the natural language processing method of this disclosure, the business parameter information is converted into request parameter information conforming to the interface technical specification based on the parameter conversion rule, and the method further includes: inputting the parameter conversion rule and the business parameter information into a second neural network model, and performing parameter conversion through the second neural network model to obtain request parameter information conforming to the interface technical specification.

[0018] According to the technical solution of this embodiment, parameter conversion is performed through a second neural network model, eliminating the need for manually writing conversion logic adapted to different interface specifications. This significantly reduces development and time costs, especially in scenarios involving multiple interfaces and high-frequency parameter conversion. Furthermore, the second neural network model can learn from a large amount of data to accurately capture the matching relationship between parameter conversion rules and interface technical specifications, thereby improving the accuracy of the parameter conversion results.

[0019] According to at least one embodiment of the natural language processing method of this disclosure, in the case that the response data returned by the interface indicates a data access failure, a response to the business query is generated based on the response data returned by the interface, comprising: performing error analysis based on the response data and the request using a third neural network model to obtain an error report; modifying the request based on the modification suggestions included in the error report; sending a modified request to the interface, the modified request being used to trigger the interface to return response data again; and generating a response to the business query based on the response data returned by the interface again.

[0020] According to the technical solution of this embodiment, the third neural network model performs intelligent error analysis and generates repair suggestions for data access failure requests, and automatically modifies the requests and retryes based on the modification suggestions, reducing manual intervention, realizing autonomous diagnosis and repair, and improving system robustness and query success rate.

[0021] According to at least one embodiment of the natural language processing method of this disclosure, after receiving the error report, the method further includes: in response to the error report not including the modification suggestion, using a preset reply content as the reply content of the business query, and / or sending an alarm prompt to the target object.

[0022] According to the technical solution of this embodiment, in the event that errors cannot be automatically corrected, timely degradation processing can be achieved by using preset response content as the response content for the business query, which can ensure that the user experience is not interrupted; by sending alarm prompts to the target object, it is convenient for the target object to quickly intervene and solve the problem, thereby improving the stability and maintainability of the natural language processing system.

[0023] According to at least one embodiment of the natural language processing method of this disclosure, when the response data returned by the interface indicates successful data access, a response to the business query is generated based on the response data returned by the interface, including: using the target type data in the response data as the response to the business query, or generating the response to the business query based on the response data using a fourth neural network model.

[0024] According to the technical solution of this embodiment, when data access is successful, the target type data in the response data can be directly used as the response content, thereby improving the efficiency of response content generation; or, the target type data can be further processed and used as the response content, thereby improving the readability of the response content.

[0025] According to at least one embodiment of the natural language processing method of this disclosure, before injecting the request parameter information into the request template to generate a request, the method further includes: in response to the inclusion of a target parameter in the request parameter information, determining whether the target parameter conforms to the format requirements; injecting the request parameter information into the request template to generate a request includes: if the target parameter conforms to the format requirements, injecting the request parameter information into the request template to generate a request.

[0026] According to the technical solution of this embodiment, a target parameter format verification step is added before request generation, and the request is generated only if the target parameters meet the format requirements. This effectively intercepts parameter inputs that do not meet the requirements, preventing subsequent interface calls from failing or malfunctioning due to format errors, improving the legality and reliability of the request, and reducing the consumption of system resources by invalid requests.

[0027] According to another aspect of this disclosure, an electronic device is provided, comprising: a memory storing a computer program; and a processor executing the computer program stored in the memory, such that the processor performs a natural language processing method according to any embodiment of this disclosure.

[0028] According to another aspect of this disclosure, a readable storage medium is provided, wherein a computer program is stored therein, which, when executed by a processor, is used to implement a natural language processing method according to any embodiment of this disclosure.

[0029] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements a natural language processing method according to any embodiment of this disclosure. Attached Figure Description

[0030] The accompanying drawings illustrate exemplary embodiments of the present disclosure and, together with the description thereof, serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0031] Figure 1 This is a flowchart illustrating a natural language processing method according to one embodiment of the present disclosure.

[0032] Figure 2 This is a schematic diagram illustrating the process of determining business parameter information according to one embodiment of the present disclosure.

[0033] Figure 3 This is a schematic diagram illustrating the process of generating request parameter information according to one embodiment of the present disclosure.

[0034] Figure 4This is a schematic diagram illustrating the process of generating request parameter information according to another embodiment of the present disclosure.

[0035] Figure 5 This is a schematic diagram illustrating the process of generating request parameter information according to another embodiment of this disclosure.

[0036] Figure 6 This is a schematic diagram illustrating the process of generating request parameter information according to another embodiment of the present disclosure.

[0037] Figure 7 This is a schematic diagram illustrating the process of generating a request according to one embodiment of the present disclosure.

[0038] Figure 8 This is a flowchart illustrating a natural language processing method according to another embodiment of the present disclosure.

[0039] Figure 9 This is a schematic structural block diagram of a natural language processing apparatus according to one embodiment of the present disclosure.

[0040] Figure 10 This is a schematic block diagram of an electronic device according to one embodiment of the present disclosure. Detailed Implementation

[0041] The present disclosure will now be described in further detail with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are for illustrative purposes only and are not intended to limit the scope of the disclosure. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the accompanying drawings.

[0042] It should be noted that, where there is no conflict, the embodiments and features described in this disclosure can be combined with each other. The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] When operational tasks such as store alerts, regional evaluation report generation, and business performance monitoring are implemented through different data sources, each data source may be developed and maintained by an independent team. This leads to significant differences in the technical specifications of the interfaces provided by different data sources. This multi-source heterogeneity makes it impossible for staff to perform operational tasks quickly and uniformly. For example, they often need to memorize the query conditions of different data sources and manually write requests that conform to the interface technical specifications in order to call the corresponding data source. This is not only cumbersome, but also prone to failure or misjudgment due to incorrect parameters or misunderstandings, thereby reducing operational efficiency.

[0044] To address this, this disclosure proposes the following technical solution: In this solution, a first neural network model parses natural language business queries into business parameter information. Based on the data source identifier in the business parameter information, it automatically matches corresponding parameter conversion rules, request templates, and interfaces. Then, based on the parameter conversion rules, it converts the business parameter information into request parameter information conforming to the interface technical specifications. Finally, it accesses the interface corresponding to the data source identifier based on the request parameter information and request template to generate a response to the business query. This effectively shields the differences in underlying data source interfaces, unifies the entry point for accessing different data sources, and automates the process. This allows users to uniformly input business queries in natural language to quickly perform various operational operations, significantly reducing the user's learning curve and improving operational efficiency.

[0045] To facilitate description and make the technical solutions of this disclosure easier to understand, the terminology of this disclosure will be explained before describing the technical solutions of this disclosure.

[0046] Natural language is the language that humans use in their daily lives.

[0047] Business queries are texts used for operational tasks such as store alerts, generating regional evaluation reports, and monitoring business indicators.

[0048] Business parameter information can be understood as information that reflects the user's intent from the front-end business level, including but not limited to the analysis object, time range, geographical range, etc.

[0049] Request parameter information can be understood as the fields, formats, and values ​​required for specific API calls from the backend technical level, such as the field names, time formats, and object encodings specified by the API.

[0050] A request template is a pre-defined framework for defining the structure and format required when sending requests to a specific interface. Request templates can include fixed parameters, placeholders, and calling conventions; by injecting specific parameters, they can generate executable, standardized requests.

[0051] Relative time refers to describing the time distance of an event relative to the current moment, such as "3 days ago" or "the last week".

[0052] Absolute time refers to the specific moment an event occurs, expressed using one or more of the following: year, month, day, hour, minute, and second, such as "October 10, 2025".

[0053] Large language models refer to deep learning models trained on large amounts of text data, enabling them to generate natural language text or understand the meaning of natural language text. It is a type of artificial intelligence technology based on deep learning.

[0054] The natural language processing method disclosed herein can automatically invoke the corresponding data source to obtain response data when a user inputs a business query in natural language form, and then generate a response to the business query based on the response data. In this disclosure, electronic devices include, but are not limited to, mobile phones, tablets, laptops, personal computers, wearable devices, and ATMs.

[0055] Figure 1 A schematic diagram illustrating the overall flow of a natural language processing method according to one embodiment of this disclosure is shown. Figure 1 The method M100 shown includes steps S110 to S160. This method can be executed by electronic devices such as mobile phones and tablets.

[0056] In step S110, in response to receiving a business query in natural language form, the business query is parsed through the first neural network model to obtain business parameter information.

[0057] The first neural network model disclosed herein can be any existing neural network model, such as deep learning-based neural network models, shallow neural network models, etc., all of which fall within the protection scope of this disclosure. For example, the first neural network model is a large language model.

[0058] For example, the input to the first neural network model is a business query, and the output is business parameter information. In this way, the trained first neural network model can intelligently and automatically determine the business parameter information.

[0059] In step S120, based on the first mapping relationship, the parameter conversion rules, request templates and interfaces corresponding to the data source identifier in the business parameter information are determined.

[0060] For example, the first mapping relationship may include a mapping between the data source identifier and the parameter conversion rule, a mapping between the data source identifier and the request template, and a mapping between the data source identifier and the interface. The specific content of the first mapping relationship can be set according to the actual situation.

[0061] In one example, the data source is identified as data source id, and the interfaces include an early warning interface, an evaluation interface, and an indicator interface. In the first mapping relationship, if the data source id prefix is ​​3, it corresponds to the early warning interface and the early warning message request template; if the data source id prefix is ​​1, it corresponds to the evaluation interface and the evaluation report request template; if the data source id prefix is ​​10, it corresponds to the indicator interface and the indicator data request template.

[0062] In step S130, based on parameter conversion rules, the business parameter information is converted into request parameter information that conforms to the interface technical specifications.

[0063] The parameter conversion rules can be set according to the interface technical specifications, and are not limited here.

[0064] In step S140, the request parameter information is injected into the request template to generate a request.

[0065] In step S150, a request is sent to the interface, which is used to trigger the interface to return response data.

[0066] Since the request is sent to the interface corresponding to the data source identifier, a successful response indicates that the data source was successfully invoked.

[0067] In step S160, the response content for the business query is generated based on the response data returned by the interface.

[0068] The natural language processing method of this disclosure can parse business queries in natural language form into business parameter information through a first neural network model. Based on the data source identifier in the business parameter information, it automatically matches corresponding parameter conversion rules, request templates, and interfaces. Then, based on the parameter conversion rules, it converts the business parameter information into request parameter information conforming to the interface technical specifications. Finally, it accesses the interface corresponding to the data source identifier based on the request parameter information and request template to generate a response to the business query. This effectively shields the differences in underlying data source interfaces, achieving a unified entry point and automated process for accessing different data sources. Users can uniformly input business queries in natural language to quickly perform various operational operations, significantly lowering the user threshold and improving operational efficiency. Furthermore, since the generation of interface access requests is automatically achieved through parameter conversion rules and request templates, even if a new data source is added or existing technical specifications are adjusted, users do not need to memorize new interface technical specifications. Adding a new data source or adjusting technical specifications can be quickly achieved by supplementing or adjusting parameter conversion rules and request templates, resulting in stronger scalability.

[0069] Regarding step S110, in some embodiments of this disclosure, it may include, for example... Figure 2 Steps S111 and S112 are shown.

[0070] In step S111, knowledge data of the domain to which the business query belongs is obtained.

[0071] As one possible implementation, the domain to which a business query belongs can be determined by the way the business query is obtained. For example, different query entry points can be pre-set for each domain, so that when a business query is received through a query entry point, the domain to which the business query belongs can be directly and quickly determined to be the domain corresponding to that query entry point.

[0072] As another possible implementation, the domain to which a business query belongs can be determined by the keywords in the business query. For example, a mapping relationship between keywords and domains can be pre-configured, so that when a business query is received, the keywords in the business query can be extracted, and the domain corresponding to the keywords can be used as the domain to which the business query belongs, thereby improving accuracy.

[0073] The knowledge data may include information related to multiple different words or phrases, which may include one or more of the following: definitions, related terms, synonyms, and near-synonyms. In addition, the knowledge data may also include rules for identifying school district housing, store ID coding systems, etc.

[0074] In step S112, the business query is parsed based on knowledge data using the first neural network model to obtain business parameter information.

[0075] For example, since business queries may be ambiguous business requests presented in natural language, these requests are characterized by strong domain dependence (e.g., needing to understand real estate terms such as "school district housing" and "finished housing") and ambiguity of elements (e.g., missing time, object, or indicator). Therefore, they need to be parsed to identify the business parameter information. A well-trained first neural network model can possess capabilities such as keyword matching, missing element completion, and structured information output. Therefore, by inputting knowledge data of the domain to which the business query belongs and the business query itself into the first neural network model, the model can automatically identify and output the corresponding business parameter information based on this knowledge data. In this embodiment, the output of the first neural network model may include knowledge data and the business query, and the output may include business parameter information.

[0076] In one example, if the business query is "Get the top 10 stores in Beijing's Haidian District with the highest sales volume of school district housing in the past 7 days", the first neural network model can parse the business type as "school district housing analysis"; the geographical scope as "Beijing" and "Haidian District"; the time dimension as "relative" with a value of "last_7d" and a granularity of "day"; and the data dimensions as "store ranking" and "top 10". Simultaneously, the first neural network model can combine the knowledge data of the business query's domain to determine that the associated indicator ID for "school district housing" is 105613, and that "store" is mapped to the business dimension "shop", and generate a school district housing label filter is_school_district=1. The first neural network model can then output the structured business parameter information shown below.

[0077] { "business_type": "School District Housing Analysis", "location": {"city": "Beijing", "district": "Haidian District"}, "time_range": {"type": "relative", "value": "last_7d", "granularity": "day"}, "data_scope": { "metric_id": "105613", "dimension": "shop", "sort": {"field": "sales_amount", "order": "desc"}, "limit": 10 } }

[0078] The natural language processing method described above introduces domain-related knowledge data during the parsing process, enabling the first neural network model to have stronger semantic understanding capabilities. Especially when faced with professional terms, polysemous words, or ambiguous expressions, it can combine domain knowledge data to perform accurate disambiguation and parameter recognition, thereby improving the accuracy and coverage of business parameter information extraction.

[0079] In some embodiments of this disclosure, the parameter conversion rules include one or more of the following: parameter conversion based on a second mapping relationship, parameter conversion by calling a target interface, and conversion of relative time to absolute time.

[0080] The natural language processing method described above defines different types of parameter conversion rules, which can meet diverse parameter conversion needs. This modular and composable rule design improves the flexibility and versatility of parameter conversion, enabling it to efficiently support parameter conversion in various business scenarios and reduce the workload of custom development.

[0081] Regarding step S130, as one possible implementation, the second mapping relationship includes at least the mapping relationship between business parameter names and request parameter names. Accordingly, step S130 may include, for example: Figure 3 Steps S131 to S133 are shown.

[0082] In step S131, based on the second mapping relationship, the target request parameter name corresponding to the target business parameter name in the business parameter information is determined.

[0083] The second mapping relationship can be set according to business needs. Different second mapping relationships can be configured for different domains. After determining the domain to which the business query belongs, the target request parameter name corresponding to the target business parameter name in the business parameter information can be determined based on the second mapping relationship corresponding to the domain to which the business query belongs, thereby improving domain adaptability.

[0084] For example, the second mapping relationship may further include a mapping relationship between business parameter names and request parameter values, a mapping relationship between business parameter values ​​and request parameter names, and a mapping relationship between business parameter values ​​and request parameter values. In the second mapping relationship, the types mapped to by different business parameter names or business parameter values ​​may be the same or different. For example, if the first business parameter name is mapped to a request parameter name, then the second business parameter name may be mapped to a request parameter name or a request parameter value.

[0085] For example, if a business parameter name in the business parameter information does not exist in the second mapping relationship, then the business parameter name may not participate in the parameter conversion, or it may be converted in other ways in the parameter conversion rules.

[0086] In step S132, the target service parameter value corresponding to the target service parameter name is used as the target request parameter value corresponding to the target request parameter name.

[0087] In step S133, the target request parameter name and the target request parameter value corresponding to the target request parameter name are used as the request parameters in the request parameter information.

[0088] The natural language processing method described above performs parameter conversion based on the second mapping relationship, achieving automatic name mapping between business parameters and request parameters, thus resolving the inconsistency between front-end natural language expressions and back-end system field naming. Furthermore, by establishing a standardized second mapping relationship, new interfaces can be quickly adapted without modifying the core logic, thereby improving development efficiency and making it suitable for complex environments with multiple data sources.

[0089] Regarding step S130, as another possible implementation, it may also include, for example... Figure 4 Steps S134 and S135 are shown.

[0090] In step S134, in response to the fact that the business parameter information includes geographic region information and object type information, the object identifier that matches the geographic region information and object type information is obtained through the target interface call.

[0091] Geographic region information describes the geographic region where the requested object is located. Object type information describes the type of object the request targets. The requested object could be a store, etc.

[0092] In one example, the geographic region information includes Haidian District, Beijing; the object type information includes stores; the target interface is a geographic service interface; and the object identifier is the store ID. Therefore, by calling the geographic service interface, the store IDs of all stores within Haidian District, Beijing can be obtained. The store ID can include the city, region, and number; for example, a store ID could be BJ_HZ_3401.

[0093] In step S135, the object identifier and object type information are used as object parameters in the request parameter information.

[0094] The natural language processing method described above, which calls the target interface to perform parameter conversion, can automatically obtain the corresponding object identifier when the business parameter information includes geographical area information and object type information. This enriches the object parameters in the request parameter information, avoids query failure due to the lack of accurate object identifiers, enhances the fault tolerance capability for fuzzy input and the completeness of data acquisition, and improves the query success rate.

[0095] Regarding step S130, as another possible implementation, it may also include, for example... Figure 5 or Figure 6 Steps S136 and S137 are shown.

[0096] In step S136, in response to the fact that the service parameter information includes a relative time range, the start and end absolute times corresponding to the relative time range are determined based on the current absolute time, and the time granularity is determined.

[0097] For example, the number of relative time ranges in the business parameter information can be one or more, and each relative time range can be a continuous range. When there are multiple relative time ranges in the business parameter information, the multiple relative time ranges may partially overlap or be completely different, and the start and end absolute times corresponding to each relative time range can be determined based on the current absolute time.

[0098] The current absolute time can be determined via system timestamp or online connection.

[0099] The time granularity can be any of the following: year, quarter, month, week, day, and hour. The time granularity can be set by default or determined based on the mapping relationship between the time granularity and the number of days contained in the relative time range. For example, if the number of days contained in the relative time range is less than or equal to 7 days and greater than 1 day, then the time granularity is determined to be day; if the number of days contained in the relative time range is less than or equal to 1 day and greater than 1 hour, then the time granularity is determined to be hour.

[0100] In one example, if the relative time range is the last 7 days and the current absolute time is 20250630, then the start absolute time (startDate) can be determined as 20250623, the end absolute time (endDate) as 20250630, and the time granularity (dateType) as days.

[0101] In step S137, the start and end absolute times and time granularity are used as time parameters in the request parameter information.

[0102] The natural language processing method described above automatically converts the relative time range expressed in natural language into absolute time that conforms to the interface technical specifications and supplements the time granularity, thereby making the time parameters in the request parameter information conform to the interface technical specifications, improving the accuracy and timeliness of time-related queries, and reducing the user's input burden.

[0103] Regarding step S130, as another implementation, it may include: inputting parameter conversion rules and business parameter information into a second neural network model, performing parameter conversion through the second neural network model, and obtaining request parameter information that conforms to the interface technical specifications.

[0104] The second neural network model disclosed herein can be any existing neural network model, such as deep learning-based neural network models, shallow neural network models, etc., all of which fall within the protection scope of this disclosure. For example, the second neural network model is a large language model.

[0105] For example, the input to the second neural network model is parameter transformation rules and business parameter information, and the output is request parameter information. In this way, the trained second neural network model can intelligently and automatically determine the request parameter information.

[0106] Continuing with the example of structured business parameter information above, in one example, the structured request parameter information obtained through the second neural network model is shown below.

[0107] { "data_source_id": "105613", "source_type": "metric", "object_type": "shop", "tech_dimension": "dimension", "object_ids": ["BJ_HZ_3401", "BJ_HZ_3402", ...], "filters": [ {"dimension": "is_school_district", "values": ["1"]} ], "time_params": { "startDate": "20250623", "endDate": "20250630", "dateType": "day }, "sort": {"field": "sales_amount", "order": "desc"}, "limit": 10 }

[0108] The natural language processing method described above uses a second neural network model for parameter conversion, eliminating the need for manually writing conversion logic adapted to different interface specifications. This significantly reduces development and time costs, especially in scenarios involving multiple interfaces and high-frequency parameter conversion. Furthermore, the second neural network model can learn from large amounts of data to accurately capture the matching relationship between parameter conversion rules and interface technical specifications, thereby improving the accuracy of the parameter conversion results.

[0109] In some embodiments of this disclosure, prior to step S140, the following may also be included: Figure 7 The step S170 is shown. Correspondingly, step S140 may include, as shown... Figure 7 The step S141 shown.

[0110] In step S170, in response to the fact that the request parameter information includes the target parameter, it is determined whether the target parameter meets the format requirements.

[0111] For example, during algorithm execution, unforeseen factors may cause the results to be random. For instance, when a neural network model is involved, the generated results may be incorrect. Therefore, checking the target parameter in the request parameter information can improve the success rate of subsequent natural language processing. If the target parameter is not included in the request parameter information, format checking is not required.

[0112] The target parameters and format requirements can be set according to needs. In one example, the target parameters include start and end absolute times and object identifiers. If the request parameter information includes start and end absolute times, it checks whether these times are within a valid time period (e.g., one year). If the request parameter information includes object identifiers, it checks whether the object identifier's format conforms to the identifier format requirements (e.g., city_region_number). This allows for validation of the start and end absolute times and object identifiers, improving the success rate of natural language processing.

[0113] In step S141, if the target parameters meet the format requirements, the request parameter information is injected into the request template to generate a request.

[0114] When there are multiple target parameters, if all target parameters meet the format requirements, the request parameter information is injected into the request template to generate a request. If any target parameter does not meet the format requirements, an error report is generated by performing error analysis based on parameter conversion rules, business parameter information, and request parameter information using a large language model. If the error report includes modification suggestions for the request parameter information, the request parameter information is modified based on the modification suggestions to obtain request parameter information that meets the format requirements. If the error report does not include modification suggestions, a preset response content can be used as the response content for the business query.

[0115] Continuing with the structured request parameter information example above, in one example, the request generated after injecting the request parameter information into the request template is shown below.

[0116] { "url": "http: / / metric.api.a.com / query", "method": "POST", "headers": {"Content-Type": "application / json"}, "body": { "indexIds": "105613", "extFilters": [ {"dimension": "is_school_district", "values": ["1"], "type": "in"}, {"dimension": "city_code", "values": ["1100"], "type": "in"} ], "startDate": "20250623", "endDate": "20250630", "sortFields": [{"field": "sales_amount", "order": "DESC"}], "limit": 10, "diagnosisQuery": { "startDate": "20250623", "endDate": "20250630" } } }

[0117] The natural language processing method described above adds a target parameter format verification step before request generation, and generates the request only if the target parameters meet the format requirements. This effectively intercepts invalid parameter inputs, preventing subsequent interface calls from failing or malfunctioning due to format errors, thus improving the legitimacy and reliability of requests and reducing the consumption of system resources by invalid requests.

[0118] Regarding step S160, in some embodiments of this disclosure, if the response data returned by the interface indicates that the data access has failed, steps S161 to S164 may be included.

[0119] In step S161, an error report is obtained by performing error analysis based on the response data and the request using a third neural network model.

[0120] For example, the response data returned by the interface may include a response status, which can be used to determine whether data access was successful. In one example, a response status of "success" indicates successful data access, while a response status of "error" indicates failed data access.

[0121] The third neural network model disclosed herein can be any existing neural network model, such as deep learning-based neural network models, shallow neural network models, etc., all of which fall within the protection scope of this disclosure. For example, the third neural network model is a large language model.

[0122] For example, the input to the third neural network model is the response data and the request content, and the output is an error report. In this way, the trained third neural network model can intelligently and automatically identify the reasons for data access failure (such as incorrect time format, incorrect dimension alias mapping, etc.) and automatically output an error report.

[0123] In step S162, in response to the error report including modification suggestions for the request, the request is modified based on the modification suggestions.

[0124] In one example, the error report looks like this.

[0125] { "status": "error", "error_code": "INVALID_TIMEFORMAT", "request_params": {original request parameters}, "error_response": {API returned an error}, "suggestion": "Please change startDate from '2025-06-23' to '20250623'" }

[0126] The error report above indicates that the suggested modification is to change startDate from '2025-06-23' to '20250623'. The format of the start time in the request is incorrect and needs to be corrected. Therefore, the start time in the request can be automatically changed from '2025-06-23' to '20250623'.

[0127] In step S163, a modified request is sent to the interface, which is used to trigger the interface to return response data again.

[0128] In step S164, the response content for the business query is generated based on the response data returned by the interface.

[0129] For example, if the response data returned by the interface indicates successful data access, the target type data in the response data can be used as the response content for the business query, or the response content for the business query can be generated based on the response data using the fourth neural network model. If the response data returned by the interface indicates failed data access, the number of failures can be accumulated. If the number of failures is less than or equal to a failure threshold, step S161 can be entered again for multiple iterations of error analysis and request modification until the number of failures exceeds the failure threshold or data access is successful. The failure threshold can be set according to requirements.

[0130] The natural language processing method described above uses a third neural network model to perform intelligent error analysis and generate repair suggestions for failed data access requests. Based on the modification suggestions, it automatically modifies the requests and retryes, reducing manual intervention, achieving autonomous diagnosis and repair, and improving system robustness and query success rate.

[0131] In some embodiments of this disclosure, after step S161, the method may further include: responding to the fact that the error report does not include modification suggestions, using the preset response content as the response content for the business query, and / or sending an alarm prompt to the target object.

[0132] The preset response content and target audience can be set according to needs.

[0133] The natural language processing method described above, when unable to automatically correct errors, achieves timely degradation processing by using preset response content as the response content for business queries, ensuring uninterrupted user experience; by sending alarm prompts to the target object, it facilitates the target object's rapid intervention and problem resolution, thereby improving the stability and maintainability of the natural language processing system.

[0134] In some implementations, after sending an alarm to the target object, the operation data of the target object and the complete execution process log of the current business query can be obtained. Based on the operation data and the complete execution process log, one or more of the following can be optimized: knowledge data of the domain to which the business query belongs, the first neural network model, the second neural network model, the third neural network model, the first mapping relationship, the data conversion rules, and the request template. This achieves a closed-loop mechanism of execution-diagnosis-optimization and improves natural language processing capabilities.

[0135] Regarding step S160, in some embodiments of this disclosure, if the response data returned by the interface indicates successful data access, step S160 may include: using the target type data in the response data as the response content for the business query, or generating the response content for the business query based on the response data using a fourth neural network model.

[0136] The target type can be set according to the actual situation. For example, the value of the "data" key in the response data can be used as the target type data.

[0137] The fourth neural network model disclosed herein can be any existing neural network model, such as deep learning-based neural network models, shallow neural network models, etc., all of which fall within the protection scope of this disclosure. For example, the fourth neural network model is a large language model. The first, second, third, and fourth neural network models can be the same model or different models.

[0138] For example, the input to the fourth neural network model is the response data, and the output is the reply content. In this way, the response data can be semantically processed and natural language generated through the fourth neural network model, making the generated reply content closer to the expression habits of natural language and improving the readability of the reply content.

[0139] In one example, the response data returned by the interface is shown below.

[0140] { "status": "success", "metadata": { "data_source": "sales revenue", "time_range": "20250623-20250630" }, "data": [ { "shop_code": "BJ_HZ_3401", "shop_name": "Zhongguancun Store", "sales_amount": 1420, "school_quality": "A+" }, ... / / TOP10 data ] }

[0141] The natural language processing method described above can directly use the target type data in the response data as the response content when the data access is successful, thereby improving the efficiency of response content generation; or, it can further process the target type data and use it as the response content, thereby improving the readability of the response content.

[0142] It should be noted that the specific values ​​mentioned above are only for illustrating the implementation of this disclosure in detail, and should not be construed as limiting the disclosure. In other examples, implementation methods, or embodiments, other values ​​may be selected based on this disclosure, and no specific limitations are made here.

[0143] Please combine Figure 8 In one example, the natural language processing method may include the following steps S201 to S220. The content related to steps S201 to S220 can be referred to the description of the above implementation method. For the sake of brevity, it will not be repeated here.

[0144] In step S201, a business query in natural language form is received.

[0145] In step S202, knowledge data of the domain to which the business query belongs is obtained.

[0146] In step S203, the business query is parsed based on knowledge data using the first neural network model to obtain business parameter information.

[0147] In step S204, based on the first mapping relationship, the parameter conversion rules, request templates and interfaces corresponding to the data source identifier in the business parameter information are determined.

[0148] In step S205, the parameter conversion rules and business parameter information are input into the second neural network model, and the parameters are converted through the second neural network model to obtain request parameter information that conforms to the interface technical specifications.

[0149] In step S206, it is determined whether the target parameter exists in the request parameter information. If it does, proceed to step S207; otherwise, proceed to step S208.

[0150] In step S207, it is determined whether the target parameter meets the format requirements. If it does, proceed to step S208; otherwise, proceed to step S218.

[0151] In step S208, the request parameter information is injected into the request template to generate a request.

[0152] In step S209, a request is sent to the interface, which is used to trigger the interface to return response data.

[0153] In step S210, the system determines whether the data access was successful based on the response data returned by the interface. If it was, the system proceeds to step S211; otherwise, it proceeds to step S212.

[0154] In step S211, the fourth neural network model generates the response content for the business query based on the response data.

[0155] In step S212, an error report is obtained by performing error analysis based on the response data and the request using a third neural network model.

[0156] In step S213, it is determined whether the error report includes modification suggestions for the request. If so, proceed to step S214; otherwise, proceed to step S217.

[0157] In step S214, the request is modified based on the modification suggestions.

[0158] In step S215, a modified request is sent to the interface, which is used to trigger the interface to return response data again.

[0159] In step S216, the response content for the business query is generated based on the response data returned by the interface.

[0160] In step S217, the preset response content is used as the response content for the business query.

[0161] In step S218, an error report is obtained by performing error analysis based on parameter conversion rules, business parameter information, and request parameter information using a large language model.

[0162] In step S219, it is determined whether the error report includes modification suggestions for the request parameter information. If so, proceed to step S220; otherwise, proceed to step S217.

[0163] In step S220, if the error report includes modification suggestions for the request parameter information, the request parameter information is modified based on the modification suggestions to obtain request parameter information that meets the format requirements.

[0164] Based on any of the above embodiments, this disclosure also provides a natural language processing apparatus.

[0165] Figure 9 This is a schematic block diagram of the structure of a natural language processing device according to one embodiment of the present disclosure.

[0166] like Figure 9 As shown, the natural language processing device includes: a parsing module 110, used to parse a business query in natural language form in response to receiving such a query, and obtain business parameter information by using a first neural network model; a first determining module 120, used to determine the parameter conversion rules, request template, and interface corresponding to the data source identifier in the business parameter information based on a first mapping relationship; a conversion module 130, used to convert the business parameter information into request parameter information conforming to the interface technical specifications based on the parameter conversion rules; a first generating module 140, used to inject the request parameter information into the request template to generate a request; a sending module 150, used to send a request to the interface, the request being used to trigger the interface to return response data; and a second generating module 160, used to generate a response to the business query based on the response data returned by the interface.

[0167] The aforementioned natural language processing device can be in the form of computer software, and each module of the aforementioned natural language processing device can be implemented through computer software modules.

[0168] In some embodiments of this disclosure, the parsing module 110 is used to: acquire knowledge data of the domain to which the business query belongs; and parse the business query based on the knowledge data using a first neural network model to obtain business parameter information.

[0169] In some embodiments of this disclosure, the parameter conversion rules include one or more of the following: parameter conversion based on a second mapping relationship, parameter conversion by calling a target interface, and conversion of relative time to absolute time.

[0170] As one possible implementation, the second mapping relationship includes at least the mapping relationship between business parameter names and request parameter names; accordingly, the conversion module 130 is used to: determine the target request parameter name corresponding to the target business parameter name in the business parameter information based on the second mapping relationship; use the target business parameter value corresponding to the target business parameter name as the target request parameter value corresponding to the target request parameter name; and use the target request parameter name and the target request parameter value corresponding to the target request parameter name as the request parameter in the request parameter information.

[0171] As another possible implementation, the conversion module 130 is also used to: in response to the business parameter information including geographic area information and object type information, obtain an object identifier that conforms to the geographic area information and object type information through a target interface call; and use the object identifier and object type information as object parameters in the request parameter information.

[0172] As another possible implementation, the conversion module 130 is also configured to: in response to the business parameter information including a relative time range, determine the start and end absolute times corresponding to the relative time range based on the current absolute time, and determine the time granularity; and use the start and end absolute times and the time granularity as time parameters in the request parameter information.

[0173] As another possible implementation, the conversion module 130 is also used to: input parameter conversion rules and business parameter information into the second neural network model, perform parameter conversion through the second neural network model, and obtain request parameter information that conforms to the interface technical specifications.

[0174] In some embodiments of this disclosure, the natural language processing apparatus further includes: a judgment module, configured to determine whether the target parameter conforms to the format requirements in response to the inclusion of the target parameter in the request parameter information; correspondingly, the first generation module 140 is configured to: inject the request parameter information into the request template and generate a request if the target parameter conforms to the format requirements.

[0175] In one possible implementation, the second generation module 160 is used to: respond to the response data returned by the interface indicating a data access failure, perform error analysis based on the response data and the request using a third neural network model to obtain an error report; respond to the error report including modification suggestions for the request, modify the request based on the modification suggestions; send a modified request to the interface, the modified request being used to trigger the interface to return response data again; and generate the response content for the business query based on the response data returned by the interface again.

[0176] In some embodiments of this disclosure, the second generation module 160 is further configured to: in response to the error report not including modification suggestions, use the preset response content as the response content for the business query, and / or send an alarm prompt to the target object.

[0177] As another possible implementation, the second generation module 160 is used to: respond to the response data returned by the interface indicating successful data access, use the target type data in the response data as the response content of the business query, or generate the response content of the business query based on the response data through the fourth neural network model.

[0178] The specific implementation process of the functions and roles of each module in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0179] The natural language processing method in the specific embodiments of this disclosure can be executed by electronic devices such as mobile phones and computers.

[0180] Therefore, based on any of the above embodiments, this disclosure also provides an electronic device that can execute the natural language processing method of any of the embodiments described above.

[0181] Figure 10 This is a schematic block diagram of an electronic device 1000 according to one embodiment of the present disclosure.

[0182] The hardware structure of electronic device 1000 can be implemented using a bus architecture. The bus architecture can include any number of interconnect buses and bridges, depending on the specific application and overall design constraints of the hardware. Bus 1100 connects various circuits including one or more processors 1200, memory 1300, and / or hardware modules. Bus 1100 can also connect various other circuits 1400 such as peripherals, voltage regulators, power management circuits, external antennas, etc. Bus 1100 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, only one connection line is used in this figure, but this does not indicate that there is only one bus or one type of bus.

[0183] For ease of explanation, certain steps of the above method are described in relation to modules. It should be understood that the corresponding module performing one or more steps of the above method may be one or more hardware modules specifically configured to perform the corresponding step, or implemented by a processor configured to perform the corresponding step, or stored in a computer-readable medium for implementation by a processor, or implemented by some combination thereof.

[0184] This disclosure also provides a readable storage medium storing a computer program that, when executed by a processor, is used to implement the methods described above. A "readable storage medium" can be any means capable of containing, storing, communicating, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples of a readable storage medium include: an electrical connection with one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable read-only memory (CDROM), etc.

[0185] This disclosure also provides a computer program product, the methods of which can be implemented wholly or partially through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented wholly or partially as a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed, all or part of the processes or functions of this disclosure are performed.

[0186] Computer programs or instructions can be stored in a readable storage medium or transferred from one readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The readable storage medium can be any available medium capable of access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video optical disc; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium can be a volatile or non-volatile storage medium, or it can include both volatile and non-volatile types of storage media.

[0187] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, apparatus, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0188] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0189] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0190] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0191] In the description of this specification, the references to terms such as "one embodiment / mode," "some embodiments / modes," "example," "specific example," or "some examples," etc., refer to specific features, structures, or characteristics described in connection with that embodiment / mode or example, which are included in at least one embodiment / mode or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment / mode or example. Moreover, the specific features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments / modes or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments / modes or examples described in this specification, as well as the features of different embodiments / modes or examples.

[0192] Those skilled in the art should understand that the above embodiments are merely for illustrating the present disclosure and are not intended to limit the scope of the disclosure. Those skilled in the art can make other changes or modifications based on the above disclosure, and these changes or modifications still fall within the scope of the present disclosure.

Claims

1. A natural language processing method, characterized in that, include: In response to receiving a business query in natural language form, the first neural network model is used to parse the business query to obtain business parameter information; Based on the first mapping relationship, determine the parameter conversion rules, request templates, and interfaces corresponding to the data source identifier in the business parameter information; Based on the parameter conversion rules, the business parameter information is converted into request parameter information that conforms to the interface technical specifications; The request parameter information is injected into the request template to generate a request; Send the request to the interface, the request being used to trigger the interface to return response data; as well as The response content for the business query is generated based on the response data returned by the interface.

2. The natural language processing method according to claim 1, characterized in that, The business query is parsed using a first neural network model to obtain business parameter information, including: Obtain knowledge data in the domain to which the business query belongs; and The business query is parsed using the first neural network model based on the knowledge data to obtain the business parameter information.

3. The natural language processing method according to claim 1, characterized in that, The parameter conversion rules include one or more of the following: parameter conversion based on the second mapping relationship, parameter conversion by calling the target interface, and conversion of relative time to absolute time.

4. The natural language processing method according to claim 3, characterized in that, The second mapping relationship includes at least the mapping relationship between business parameter names and request parameter names; Based on the parameter conversion rules, the business parameter information is converted into request parameter information that conforms to the interface technical specifications, including: Based on the second mapping relationship, determine the target request parameter name corresponding to the target business parameter name in the business parameter information; The target service parameter value corresponding to the target service parameter name is used as the target request parameter value corresponding to the target request parameter name; and The target request parameter name and the target request parameter value corresponding to the target request parameter name are used as the request parameters in the request parameter information.

5. The natural language processing method according to claim 4, characterized in that, Based on the parameter conversion rules, converting the business parameter information into request parameter information that conforms to the interface technical specifications further includes: In response to the fact that the business parameter information includes geographic region information and object type information, an object identifier matching the geographic region information and the object type information is obtained through a target interface call; and The object identifier and the object type information are used as object parameters in the request parameter information.

6. The natural language processing method according to claim 4 or 5, characterized in that, Based on the parameter conversion rules, converting the business parameter information into request parameter information that conforms to the interface technical specifications further includes: In response to the fact that the service parameter information includes a relative time range, the start and end absolute times corresponding to the relative time range are determined based on the current absolute time, and the time granularity is determined; and The start and end absolute times and the time granularity are used as time parameters in the request parameter information.

7. The natural language processing method according to claim 1, characterized in that, Based on the parameter conversion rules, the business parameter information is converted into request parameter information that conforms to the interface technical specifications, including: The parameter conversion rules and the business parameter information are input into the second neural network model, and the parameters are converted through the second neural network model to obtain request parameter information that conforms to the interface technical specifications.

8. The natural language processing method according to claim 1, characterized in that, If the response data returned by the interface indicates a data access failure, a response to the business query is generated based on the response data returned by the interface, including: An error report is generated by performing error analysis based on the response data and the request using a third neural network model. In response to the error report including modification suggestions for the request, the request is modified based on the modification suggestions; A modified request is sent to the interface, the modified request being used to trigger the interface to return response data again; and The response content for the business query is generated based on the response data returned by the interface.

9. An electronic device, characterized in that, include: Memory, the memory storing computer programs; as well as A processor that executes a computer program stored in the memory, causing the processor to perform the natural language processing method according to any one of claims 1 to 8.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the natural language processing method according to any one of claims 1 to 8.

Images